WP 2002-18
June 2002
Working Paper
Department of Applied Economics and Management
Cornell University, Ithaca, New York 14853-7801 USA
EDUCATION, EMPOWERMENT,
AND GENDER INEQUALITIES
Ravi Kanbur
It is the Policy of Cornell University actively to support equality of
educational and employment opportunity. No person shall be denied
admission to any educational program or activity or be denied
employment on the basis of any legally prohibited discrimination
involving, but not limited to, such factors as race, color, creed, religion,
national or ethnic origin, sex, age or handicap.
The University is
committed to the maintenance of affirmative action programs which will
assure the continuation of such equality of opportunity.
2
EDUCATION, EMPOWERMENT AND GENDER INEQUALITIES
by
Ravi Kanbur*
Cornell University
www.people.cornell.edu/pages/sk145
sk145@cornell.edu
June, 2002
Contents
1.
2.
3.
4.
5.
Introduction
Are Gender Inequalities “Large”?
Do Gender Inequalities Inhibit Efficiency and Growth?
Is Inequality of Power the Fundamental Gender Inequality?
Conclusion
Abstract
This paper considers a seeming disconnect between the consensus in policy circles that
reducing gender inequalities is to be prioritized in strategies for reducing inequality and
poverty, and a view in mainstream economics (and in some policy circles) that gender
inequalities are overemphasized. This latter view is not stated openly, it being politically
incorrect to do so, but is nevertheless present. In specific terms, there is a sense that
gender inequalities are not large relative to other types of inequalities, that the evidence
on the consequences of gender inequality for economic growth is weak, and that in any
event inequality of power is not something that should receive policy priority over
conventional economic interventions. This paper takes these positions seriously, and
argues that on some readings the narrowly economic evidence does indeed support them,
but that to some extent this is an issue with the economic evidence and with its
interpretation. A reexamination of the evidence and the arguments suggests a number of
directions for research and analysis in exploring the economics of gender inequalities.
*
T.H. Lee Professor of World Affairs and Professor of Economics, Cornell University. Paper originally
prepared for presentation to the Annual World Bank Conference on Development Economics, Washington,
D.C., April 29-30, 2002. I am grateful to my two discussants, Christina Paxson and Jungyoll Yun, to two
anonymous referees, and to participants at the conference for their comments.
1. Introduction
“Gender inequality hurts all members of society, not just girls and women.”
--World Bank, http://www.worldbank.org/developmentnews/stories/html/120701a.htm
“Are we focusing on women because we believe that gender is a good category for
addressing the inequalities in the world? That is certainly false by an order of magnitude,
maybe two.”
--Anonymous reviewer of the initial proposal for this paper.
As with trade liberalization, gender equity is part of the World Bank’s mantra on
equitable development and poverty reduction. Unlike trade liberalization, gender equity
is the part of the mantra that I personally welcome and repeat. But apparently, not
everyone is convinced. The stark position taken above by an anonymous reviewer of an
earlier proposal for this paper could be easily caricatured and dismissed, but that would
be a mistake. 1 Although such positions are rarely stated in public these days, because it is
politically incorrect to do so, they are more widespread in mainstream economic thinking
than is commonly realized. And, what is more, they are intimately related to the
economic evidence on gender inequality, and to the frameworks for interpreting them.
What might underlie the disquiet about the emphasis being given to gender
inequalities? I identify three possible strands in this paper. First, there is a sense that
gender inequalities, in education and in other variables, are not as large as they are made
out to be, in comparison with inequalities along other dimensions such as country of
residence. Second, there is the argument that the macroeconomic evidence for the
beneficial effects on growth of reducing gender inequalities, again in education for
example, is weak. Third, there is a view that that to the extent that gender inequalities
need to be addressed, the focus should be on economic and social inequalities, such as
inequality in consumption or in education, and not on inequalities in power.
Why do such views persist at the core of mainstream economic thinking? I
believe the answer is that the narrowly economic evidence can be, and is, read as giving
support to these positions. At least, it is seen as not supporting prioritization of reducing
gender inequalities as strongly as the policy conventional wisdom now seems to
advocate. 2 This paper argues that there is some truth to this characterization of the
1
My original proposal, entitled “Education and the Empowerment of Women,” was for a paper that
attempted to integrate intra-household bargaining models (for example of the type in Ghosh and Kanbur,
2002) and political economy models of policy determination, in the context of educational choices and
expenditures. But when I read the general arguments being advanced in the review I set about looking at
the evidence on gender inequalities in detail, and the nature of the paper changed (the modeling exercise
will have to wait for another occasion).
2
The most recent statement of the policy conventional wisdom is World Bank (2001).
currently available economic evidence. However, this is a reflection on the nature of that
evidence. It is important for economists and policy analysts to understand these concerns
and to address them where possible—in narrowly economic terms as well as in broader
disciplinary contexts.
The outline of the paper is as follows. Section 2 takes up the exact sense in which
gender inequalities are “large”, by considering different ways of defining “large” and
through different comparisons of inequalities in various outcomes. Throughout the
paper, there will be a focus on gender inequalities in education, because of the
prominence it receives in the literature, but other aspects such as consumption will also
be considered. Section 3 looks at the evidence on gender inequality and economic
growth, and finds that in conventional macro economic studies, causality is not as
strongly established as one would like. Section 4 argues that the dominant household
model in mainstream economics, the unitary household with unique and given
preferences, tends by its nature to emphasize economic inequalities over political ones.
Section 5 concludes the paper with a discussion of the implications of the disconnect
between the policy consensus and the narrowly economic evidence and argument, and
highlights the interesting research questions that arise as a result.
2. Are Gender Inequalities “Large”?
There is considerable evidence that, by and large (with exceptions, of course), the
average achievements of women in consumption, health and education are lower than
those of men. 3 But, these differences are invariably smaller than, for example,
differences across developed and developing countries, between rural and urban residents
within a country, or between the top and bottom quintiles within a country. How are we
then to address the question of whether or not gender inequalities are “large”?
There are both conceptual and empirical issues. Conceptually, a common enough
approach is the following. Consider any variable measuring individual attainment, the
inequality of which is of interest. This could be educational attainment, or consumption,
for example. The inequality in the distribution of this variable across individuals can be
measured using any one of a number of standard inequality indices. Suppose now that
individuals also have other characteristics such as gender, location, or age. Each
characteristic divides individuals into mutually exclusive and exhaustive groups. The
overall distribution of the variable of interest can now be seen as composed of subdistributions for each group. Inequality between the groups can then be related to the
differences in the means between the groups, with inequality within groups reflected in
the spread of the variable around each group’s average.
If gender is the characteristic in question, then gender inequality can be measured
simply in terms of the differences in the mean of the variable for men and for women
(this is also an estimate of the difference in the expected value of the variable conditional
upon being a member of one group or another). Another method is to “decompose”
3
See, for example, the documentation in Mammen and Paxson (2000).
2
overall inequality into its “between group and “within group” components following
standard techniques, and to use the between group component as the measure of gender
inequality. 4 Yet another method is to regress the variable against a gender dummy and
to use the percentage of variation explained as a measure of the degree of gender
inequality. Any of these measures can then be compared to the corresponding measure
for another characteristic, for example location, to draw a conclusion about whether
gender inequality is “large” relative to other groupings in society.
All three of the above measures are used in the literature to argue the case for the
importance, or otherwise, or gender inequality. 5 But it is important to make sure that like
is being compared with like. The most common way in which this requirement is
violated is when comparing the measures for characteristics that do not have the same
number of categories. For example, the problem arises if gender, with two categories, is
compared with location (e.g. counties in international comparisons, or provinces for
comparisons within a country), which has more categories. For a start, it is not clear how
mean differences between two groups are to be compared to mean difference between
more than two groups. The second measure above does allow comparisons in this case,
but as the number of categories increase the within group component could decrease.
There is of course no guarantee of this, since it depends on the exact nature of the
groupings. But it certainly happens in the limit--when the number of categories is the
same as the number of individuals, only the between group component remains and the
within group component is zero.
Consider therefore whether the gender dimension of educational inequality is
“large” when comparing gender with wealth (or income), which is often done, with the
conclusion that wealth “accounts for” more inequality than does gender. But wealth, in
these analyses, is an individual level variable—in other words, there are as many
categories for this characteristic as there are individuals. It is perhaps not surprising,
then, that gender-based inequality in education is not “large” compared to wealth-based
inequality in education. A fairer comparison would be to treat wealth also as two
categories—high and low. But even here, there is no natural dividing between high and
low wealth (in the way that there is a natural division for gender), although above and
below the population mean might be one option. Clearly, wealth inequality in education
would loom less large if this were done, but the actual numerical values will vary across
empirical contexts.
The above discussion for inequality has a natural extension to poverty. Given a
critical cut off (the “poverty line”) for the variable in question (educational attainment or
consumption), a number of poverty indices can be used to describe the lower end of the
distribution. Of particular interest are the FGT family of indices, which can be
decomposed across characteristics that divide individuals into mutually exclusive and
4
See Shorrocks (1984).
For example, the work of Filmer (1999) is sometimes used to argue that when wealth, gender and
residence gaps are used simultaneously to explain variations in social outcomes, not much is left for gender
to explain once the other two variables are in play.
5
3
exhaustive groups as before. 6 In this case, each FGT index can be written as a weighted
sum of group FGT’s, each weight being the population share of the group in question.
The analog to the question of whether gender inequality is high is now whether
differences across gender groups contribute “ a lot” to overall poverty. The corresponding
thought experiment might be to equalize means across gender groups and ask, how much
does this change poverty, compared to the same exercise for another characteristic with
the same number of categories.
Such an exercise has not been done in the literature, to my knowledge. However,
the following thought experiment has been analyzed. 7 Consider a small redistribution
from one category to another, taking small amounts from each individual in one group
and transferring the proceeds to individuals in the other group. Such marginal
redistributions may in any case be more policy relevant than complete elimination of
inequality between groups in one fell swoop. What is the impact on poverty? The answer
clearly depends on the details of the redistribution. For the case where subtractions and
additions are additive, a particularly clear result is available for the FGT family. For
FGT(a), where ‘a’ is the famous “poverty aversion” parameter, such a redistribution
reduces poverty in proportion to the difference between FGT(a-1) for the two groups.
Thus, from this perspective, absolute differences in FGT(a-1) for men and women
become a measure of how “large” gender inequality is. This difference is to be compared
to the corresponding difference for the two groups defined by another characteristic such
as location (e.g. urban/rural), age (young/old) or employment status
(employed/unemployed). 8
The thought experiments above have a key feature—they keep the overall mean
of the variable in question constant. This accords well with the “constant budget
comparisons” principle of modern public finance analysis when the variable in question
is consumption or income. But thinking of the variable as educational attainment raises
questions about the constant budget requirement, unless it is assumed that the unit cost of
educational attainment is constant. More generally, there is the question of the actual
administrative and economic costs of the redistribution attempted in the thought
experiment. These costs are unlikely to be the same across different categories such as
gender, age or location. Thus using the “thought experiment impact” of redistribution
across categories, as a measure of the quantitative magnitude of the inequality across
these categories, is not complete without an assessment of the costs of these
redistributions. Such comparative assessments are not found in the literature. And yet,
without such analysis, any answer to the question of whether gender inequalities are
“large” will not have adequate conceptual foundation.
So much for the conceptualization of whether gender inequality is “large”. In
empirical implementation, a fundamental problem presents itself. This is that, for the
6
See Foster, Greer and Thorbecke (1984).
Kanbur (1987).
8
The analysis can be extended from giving guidance on marginal redistributions to a characterization of the
fully optimal redistribution, using this or that category across which to redistribute. See Ravallion and Chao
(1989) and Ravallion and Sen (1994) for such “optimal targeting” calculations for geographical and for
land holding categories, respectively.
7
4
standard consumption based measures of poverty, we cannot identify individual values
through our standard household survey data sets. While education can be measured
directly as an individual attainment, consumption data is collected at the household level,
and the way of generating individual level consumption is to assume that allocation is
equal distribution across the members of the household (or, in a few studies, according to
adult equivalent scales). Thus the most commonly used variable in distributional analysis
simply cannot address gender differences per se. Rather, differences in gender are forced
to be differences across households in which the genders live. The same is true of age.
The household level characteristics include, of course, location and the average
consumption or wealth of the households. Is it any wonder then that gender does not
explain individual outcomes in consumption over and above wealth of the household or
average consumption of the household, or averages for region or country of the
household?
There is a well-known literature on comparing “female headed households” with
the average household, to try and get at gender differences. Interesting as these
comparisons are in their own terms (for example in highlighting the plight of widows in
India), it should be clear that the comparison cannot conceptually be used to pronounce
on whether gender difference are “small” or “large”. 9 There is also a small literature on
genuinely trying to identify individual consumption by focusing on food intake. One
study, which used calorie intake (adjusted for needs) and calculated the understatement of
true inequality and poverty compared to standard procedures that ignore intra-household
inequality, came up with a figure of 30-40 %. But this was for all aspects of intrahousehold inequality, not just gender differences. 10
The prospects for getting true measures of individual level consump tion are not
bright. The problems are conceptual as well as empirical. The individual food
consumption data methods have their own problems. Once we get beyond food, it is not
clear how precisely to allocate household public goods consumption across individuals,
especially if preferences differ across individuals. Thus, we are stuck with a situation
that our standard “headline” measures of consumption based poverty assume no gender
difference within the household, and there is no way of assessing how serious an
assumption this is within the same framework. Not surprisingly, other variables,
information on which can indeed be collected at the individual level, will always remain
important in the study of gender differences. These variables include education, health,
income streams and time use. And assessing whether gender inequality is “large” in
terms of these variables takes us back to the issues raised at the start of this section, and
underscores the importance of addressing the conceptual issues raised there in specific
empirical contexts.
9
In any event, the literature is inconclusive. For example, Buvinic and Gupta (1997) find a preponderance
of cases where female headed households poorer than average, but Quisumbing, Haddad, and Pena (2000)
find a significant difference in very few of the cases they consider.
10
See Haddad and Kanbur (1990). But other nutrition based studies are inconclusive on whether nutrition
within the household is unequally distributed relative to need—for example, Appleton and Collier (1995).
5
3. Do Gender Inequalities Inhibit Efficiency and Growth?
Whether gender inequalities are “large” is important in understanding whether
gender based redistribution could have a major impact on overall inequality and poverty,
in particular compared to inequalities in other dimensions. But this may also be
important for another reason—if it can be showed that gender inequality impedes overall
efficiency and growth, either directly or because of its contribution to overall inequality.
In particular, the gender dimension of education inequality is often emphasized as
holding back economic growth.
The most striking feature of the literature encompassing inequality, education,
gender and growth is the strong disconnect between the theoretical and micro-empirical
studies on the one hand, and the macro-empirical studies on the other. The former
invariably produce arguments or evidence for why education inequality and inequality in
general, and gender inequality in particular, can impede efficiency and growth. The latter
set of studies tends to be far more agnostic, if not directly contradicting the micro studies.
Focusing on one set versus the other could give a totally different picture on gender
inequality, efficiency and growth.
Start with inequality in general. A whole host of theoretical studies in the last
fifteen years suggest that, contrary to the earlier conventional wisdom that inequality
helped growth because the rich saved a higher proportion of their income than the poor,
high inequality in a general sense can be detrimental to efficiency and growth. This
literature has been surveyed in Kanbur (2000) and in Kanbur and Lustig (2000), and there
is no need to go into detail here. 11 Suffice it to say that the basic structure of the
argument takes off from a second best world where, for example, incentive compatibility
constraints bite. Redistribution can then release these constraints, thereby permitting
welfare gains, even perhaps Pareto improvements. 12 Another line of argument follows
from political economy effects on policy, it being shown that a more equal distribution of
endowments leads to a more pro-growth policy being chosen by the political institutions.
In stark contrast to the tone of the theoretical literature, the empirical literature is
much more circumspect. The “reverse Kuznets effect” literature, which tries to explain
growth in a cross-section of countries as a function of the standard variables plus
inequality at the start of the growth period, started off with a strong push in the direction
of a negative relationship between inequality and growth. 13 However, recent papers have
been less clear cut, some even suggesting that the relationship is positive, which is back
to the old conventional wisdom. 14 No doubt the debate will continue as other papers find
that results in either direction are “fragile”. 15 The fact that the empirical literature is
11
See also Aghion, Caroli and Garcia-Penalosa (1999), or Benhabib and Rustichini (1996).
Such an argument is developed by Hoff and Lyon (1995), for example.
13
Alesina and Rodrik (1994), Persson and Tabellini (1994), and Alesina and Perotti (1996) are a selection
of papers in this vein.
14
Forbes (2000) finds a positive and significant association between inequality and growth, while Banerjee
and Duflo (2000) find that the growth rate is an inverted-U shape function of changes in inequality.
Arguello (2002) finds that “there is virtually no panel estimation evidence here of a negative correlation
between inequality and growth.” See also Li and Zou (1998).
15
See for example the recent paper by Morrissey, Mbabazi and Milner (2002).
12
6
inconclusive is surely related to the well-known problems of data and method, well
known from the old Kuznets curve literature, which itself found no relationship between
inequality and per capita GNP in cross section econometrics. 16 While the inequality data
set has improved over the last two decades, major problems of comparability and quality
still remain. 17 Methodologically, the central problem in establishing causality is in
identifying exogenous movements of the inequality variable that are not capturing other
differences to a large extent. 18
Turning now to gender inequality, there is considerable micro level evidence that
gender asymmetries of various sorts lead to inefficiency. Typically, these studies argue
that women are constrained from efficient use of certain inputs like credit, or inefficiently
low supply of effort because of labor market discrimination, which leads to inefficiently
excessive exploitation of other inputs, like common property resources. Equalization, in
the sense of lifting these constraints on women can thus be shown to increase
efficiency. 19 More general arguments flow from the strongly established effect of female
education on fertility and a (more controversial) causal relationship of low fertility to
growth. 20 But again, finding effects at the macro level, for example finding a strong
relationship between gender inequality in education and growth, proves to be more
difficult.
Before turning to the effect of gender inequality on growth, it should be noted that
there is a strong disconnect between micro and macro results on education and growth in
general. At the micro level, human capital theory underpins a strong relationship
between earnings and education, a relationship that is one of the more strongly
established in micro applied economics.21 Of course, this relationship applies equally to
men and women. Some years ago, this micro relationship was also argued to be present
in the macro data, showing that as average education levels increased, average income
levels and income growth rates also rose. But research in the last decade has left the
position much less certain, particularly for developing countries. There was a
tremendous expansion of education as officially measured in the decades after 1960.
However, the effects on growth have proved difficult to detect (after all, as education
levels increased from the 1960s to the 1990s, growth rates moved in the opposite
direction for many countries, especially in Africa). 22
16
For an early critique of studies that claimed to find a Kuznets curve, see Anand and Kanbur (1993). For
more recent critiques, using expanded data sets, see, for example, Li, Squire and Zou (1998).
17
A critique focusing simply on the OECD data in the newly available data sets is provided by Atkinson
and Brandolini (2001). The problems apply with greater force to developing country data sets.
18
In a recent paper, for example, Easterly (2001a) claims to have found a novel instrument—settler
mortality in newly settled colonies.
19
See the papers by Tzannatos (1999) and Ilahi (2000).
20
Jejeebhoy (1995), Klasen (1999).
21
For a recent review of the empirical literature on education and income, see Case (2001a). The micro
evidence on returns to education for girls is highlighted in Schultz (2001).
22
Easterly (2001) presents a formidable critique of the methods and conclusion of the education and growth
literature based on the work of Pritchett (1997), Klenow and Rodriguez-Clare (1997) and Judson (1996),
Benhabib and Spiegel (1994), among others. However, for a recent counter argument see Cohen and Soto
(2001).
7
Perhaps one reason for the lack of a relationship between levels of education and
growth is that gender inequality in education was increasing during this period. In fact,
this inequality, as measured by the differential enrollment rates, also generally declined,
although its levels remain high. Of course, this could still be consistent with a positive
contribution of education equity to growth, and we need to look at the analysis more
carefully. In doing so, we should be careful to distinguish between the effect of simply
increasing female education holding male education constant, and a genuine inequality
effect over and above the level effect. In any event, as documented in World Bank
(2001), despite some studies that argue for a positive effect of gender education equality
on growth, the picture emerging from cross-country regression analysis is decidedly
mixed. 23 But this should not be surprising, given the inconclusive nature of the broader
literature on education and growth and on inequality and growth. Indeed, it would be
surprising if it were otherwise.
4. Is Inequality of Power the Fundamental Inequality?
Whatever one’s view of whether gender differences are “large”, and on their
consequences for inequality, poverty and growth, why are female achievements (in
education, but also along other dimensions) lower than those of males?
One way to approach this is to consider the interaction between household level
processes, and processes and parameters that are outside the household, in the
community, the country and the world. Consider two models of the household, “unitary”
and “bargaining”. In the unitary model, the household acts as if it has a single and given
set of preferences and chooses its actions to satisfy these preferences given the constraint
set. 24 In this setting, as the outside parameters become more unfavorable to women—for
example, as gender discrimination in the labor market increases, or as the schooling that
girls receive worsens, the household will rationally deploy its labor accordingly. It will
send its women out to work less and send its girls out to schools less as well. But any
given level of total household resources will still be divided in the same way as before,
according to the household preference function.
In the bargaining model, members of the household (for simplicity, the man and
the woman) have outside options, but are in a household for various reasons, among
which are various types of joint production and scale economies. 25 The issue of
distribution of these benefits, as viewed by individual preferences, is an open one in the
bargaining model. It depends on, among other things, the outside options facing
23
Barro and Lee (1994) argued that gender inequality actually improved growth. Dollar and Gatti (1999)
and Klasen (1999) are the current studies that argue for a significant effect of gender inequality in
education on economic growth. The exercises in Knowles, Lorgelly and Owen (2000) show how sensitive
the conclusions can be to different specifications.
24
A standard application of this framework to agricultural household is seen in Singh, Squire and Strauss
(1986).
25
One of the first attempts to formalize this set up is found in Manser and Brown (1980). For more recent
modeling in this vein, see Kanbur and Haddad (1994) and Ghosh and Kanbur (2002).
8
individuals. But these outside options are themselves determined by parameters and
processes outside the household.
With this framework, there is an intricate and interacting relationship between
inequality of “power” and inequality of outcomes for men and women. But it is different
for the two views of how a household operates. Thus with unitary households, external
moves to improve education for women, by removing discriminatory practices, will
improve the position of the household as a whole, assuming its preferences were such as
to want to send girls to school. But the same applies to credit constraints, land inequality,
and a plethora of other external factors that constrain households independently of a
gender dimension. However, with a bargaining view of the household, outside
parameters that affect outside options of men and women have a deep impact on the
distribution of the gains from household activity, including the distribution of
consumption, expenditures on education, health care, time allocation and so on.
There are two senses in which power is being exercised here. One is in the
determination of the outside parameters, to the extent that they are affected by political
processes and social norms. This is the same for the unitary and the bargaining models.
The other is in the determination of outcomes within the household. This is radically
different for the two models. In particular, notice that in the unitary model women and
men will be equally interested in changing outside parameters in so far as the change
benefits their household’s total resources (which will then be divided according to the
household preferences). Put another way, women will not necessarily be more interested
in reforming discriminatory land ownership regulations than, say, improving overall
credit to the region in which they live. Of course changing these overall policies and
structures will pit different types of households, and perhaps rich versus poor households,
against each other. But there will not be a gender dimension to the politics, certainly not
within the household by definition, but also not outside the household.
However, the bargaining model leads to a gender dimension of politics inside and
outside the household. Outside options determine bargaining power within households,
and bargaining power determines outcomes for men and women in terms of individual
outcomes within the household. These are the two precise senses in which inequality of
power is the fundamental inequality explaining outcomes—first, intrahousehold power
given the outside parameters, second the power to influence the political process that
affects the outside parameters. Notice that, on the second, with a bargaining model there
is a clear gender dimension to politics outside the household as well.
Now, there is a sense in which, even in the bargaining model, it can be argued that
direct political power is not as fundamental. This is found in an oft heard argument, that
it is better to improve women’s education and their labor market position rather than, for
example, to improve their capacity to organize themselves into a political force to change
these parameters for themselves. The reasoning is that improving education opportunities
and other economic opportunities for women will strengthen their bargaining power
within the household, which will further improve their outcomes and resources. These
improved resources will then provide the base for women to organize themselves.
Without economic resources to start with, organization to change outside parameters will
9
be weak. In this sense therefore, political empowerment must take second place to
economic advancement for women, which will empower them within the household and
then, eventually, outside the household.
This is a powerful argument, but it can be countered to some extent, on the
grounds that we cannot assume that increased education for women, and reduced
discrimination in factor markets, will happen just like that. Such changes are themselves
outcomes of political processes, and unless the preferences of women are given weight in
these processes, either directly or through “enlightened” intermediaries, they will not
happen, or not happen as fast and not in quite the best way for women. While
enlightened intermediaries are always welcome, and there may be long periods when they
are all that are on the scene, the objective must always be to increase the voice and power
of women in political and social processes outside the household. The difficulty is partly
that we do not have any idea of the relative costs, per unit of gain in outcomes for
women, of the economic versus empowerment alternatives—supporting access to credit
for women versus supporting women’s organizations, for example. In the absence of
such specific evidence, there is polarization on which tack to take, with the standard
compromise reached in policy syntheses, that we should do both. 26
But the basic argument on inequality of power between the genders being
fundamental, cannot get off the ground if the unitary model rules the roost, which it still
does among mainstream economists. This is despite many studies that purport to show,
for example, that the income pooling hypothesis, a key implication of the unitary model,
is not supported by the data. According to this hypothesis, what should matter to
household expenditure patterns are the total resources of the household, not who brings in
those resources. Again and again, it is shown that who brings in the income does matter
to the allocation of household resources. 27 But mainstream economists, as evidenced by
the bulk of ongoing research on demand patterns, on general policy analysis, and by what
is taught in basic courses, still continue to be unitarist by instinct, raising methodological
issues with these studies. In particular, the standard argument is that the same factors
might affect the composition of income as, for example, relative expenditure on women’s
clothing (because a woman who goes out to work will need to spend more on work
clothes). 28 Even though seven years ago, in light of the mounting evidence a manifesto
was issued to “shift the burden of proof” to those who would argue in favor of the unitary
26
As seen, for example, in World Bank (2001). Jungyoll Hun, in his discussant’s comments, argued that in
Korea the improvement in women’s education had not directly led to the improvement of their
intrahousehold bargaining power. He also highlighted the key role of social norms and values—it was only
in 1990, as the result of political organization by women’s groups, that key legal provisions were
introduced that gave a woman the right to inherit their parent’s and husband’s property. While this
observation does not resolve the “economic versus political intervention” argument, it should certainly be
kept in mind in assessing the relative efficacy of the two types of interventions.
27
See the impressive compilation in Appendix 4 of World Bank (2001). The studies include: Browning and
Chiappori (1998), Quisumbing and Maluccio (1999), Lundberg, Pollak and Wales (1997), Haddad and
Hoddinott (1994) and Thomas (1997).
28
As Chris Paxson pointed out in her discussant’s comments, some recent papers, by Duflo (2000),
Chattopadhyay and Duflo (2001) and Case (2001b), are not subject to this type of criticism, and all provide
strong evidence against the unitary model.
10
model, this shift is not really seen in economics or even in development economics. 29 As
a result, interventions that directly address inequality of power do not get as strong a
support as they might from mainstream economics.
5. Conclusion
This paper has argued that the narrowly economic evidence can indeed be read as
supporting the view that gender inequalities (in education in other variables) are not
large, that they do not necessarily impede economic growth, and that in any case
addressing gender inequalities of power should receive less priority than more
conventional economic interventions. While those who hold these views could be
dismissed as being in a minority, this would be a mistake. Despite the impressive
synthesis represented by World Bank (2001), such views are more widespread than
commonly realized, and are in any case intimately connected to the nature of economic
evidence and the framework for interpreting them. Taking the views seriously leads to an
interesting research and data collection agenda even in terms of conventional economic
analysis.
Comparing the contribution of gender inequalities to overall inequality, with other
divisions such as age or location, poses a basic conceptual problem because while gender
has only two categories, other divisions tend to have more than two. There is thus a
natural tendency for the contribution of gender to be understated. Correcting for this
understatement raises interesting analytical questions—for example, how does the
between-group component of decomposable inequality indices behave as the number of
groups increases? Clearly, any general conclusions can only be reached in terms of the
mathematical expectation of this component across all possible two-group divisions,
versus all possible three-group divisions, four-group divisions, etc. I am not aware of
such an analysis in the literature, but it is essential as the first step to an empirical
assessment of whether gender inequalities are “large” relative to inequalities along other
dimensions.
However, more important than this technical question is that the costs of
alternative forms of inequality reduction—gender based, country of residence based,
rural-urban based, wealth-based, etc—need to be incorporated into the analysis before
any conclusion can be reached that an inequality along a particular dimension is “large.”
If by “large” is meant “that which policy should focus on as a priority”, the cost side of
the intervention has to be analyzed, and that depends on the details of the policy
instruments being used. Such specific analysis will in the end prove more productive than
a debate on whether gender inequalities are large in the abstract, seen purely as a
measurement issue.
Despite the evidence and arguments marshaled in World Bank (2001), there is not
a groundswell of consensus in mainstream economics on the macroeconomic evidence
for the positive benefits of gender equality for growth. Partly this is because the cross29
See Alderman et. al. (1995).
11
country regressions based evidence on causality from inequality in general to growth in
particular is decidedly mixed. Kanbur and Lustig (2000) noted that “the jury is still out”
and, since then, studies have continued to appear that have supported one line or the
other. It is also partly because, given that gender inequality in education is a key focus,
the macroeconomic empirical literature on growth and education (in contrast to the
microeconomic literature on education and income) is itself deeply inconclusive. These
strands of the literature set the stage for skepticism on discovering strong causal
connections between gender inequality and growth as the literature now stands. There is
no alternative here but to persevere, at every level of the literature, with finding
persuasive instrumental variables that can convince a skeptical profession that the
causality issue really has been tackled, as well as continuing to improve the quality of the
data sets.
After two decades of mounting evidence that indicates violations of the
predictions of the unitary model of the household, mainstream economics still continues
to use it as the workhorse model, and to teach it to its students as the dominant view of
the profession. It is argued here that so long as the unitary model dominates economics
teaching and discourse, inequalities of power will naturally get secondary importance in
comparison with standard economic (and social) interventions. This is something that
needs to be tackled at the core of mainstream economics through yet more evidence on
violations of the unitary model assumption, but also through the increased deployment of
non-unitary approaches, in modeling and in empirical analysis, to “conventional” topics
such as optimal taxation policy, consequences of trade for income distribution,
composition of public expenditure, etc. 30 The intrahousehold development economics
literature is, for the moment, ahead of the curve.
Thus the seemingly contrarian views on gender inequality link closely to the
current state of development economics. Even within its own terms, taking these views
seriously leads to very interesting lines of research. But, in conclusion, let us note that
there is a vast realm of evidence produced by other disciplines and other methodologies,
evidence that supports the view that gender inequalities are large, that they impede
efficiency, and that inequality of power is fundamental. Such evidence is partially
surveyed in World Bank (2001). But this evidence is typically qualitative in nature,
differing in methods and interpretation from the typically quantitative approaches in
economics. Taking qualitative approaches seriously, and integrating them with
quantitative approaches, is another line of enquiry that stands out for development
economics as it grapples with the fundamental issue of gender inequalities, their causes,
and their consequences. 31
30
See, for example, the exercises in Kanbur and Haddad (1994), Haddad and Kanbur (1993), and Basu
(2001).
31
For an assessment of complementarities between qualitative and quantitative methods, and ways of
combining them, see Kanbur (2001). On development economics and other social science disciplines, see
Kanbur (2002).
12
References
Aghion, Philippe, Eve Caroli and Cecilia Garcia-Penalosa. 1999. “Inequality and
Economic growth: the Perspective of the New Growth Theories.” Journal of Economic
Literature, 37(4):1615-1660.
Alderman, H, P.A. Chiappori, L. Haddad, J. Hoddinott and R. Kanbur. 1995. “Unitary
versus Collective Models of the Household: Is it Time to Shift the Burden of Proof?”
World Bank Research Observer 10:1-19.
Alesina, Alberto and Dani Rodrik. 1994. “Distributive Politics and Economic Growth.”
Quarterly Journal of Economics, 109(2):465-490.
Alesina, Alberto and Roberto Perotti. 1996. “Income Distribution, Political Stability and
Investment.” European Economic Review, 40(6):1203-1228.
Anand, Sudhir and Ravi Kanbur. 1993. “Inequality and Development: A Critique.”
Journal of Development Economics, 41:19-43.
Appleton, Simon and Paul Collier. 1995. “On Gender Targeting of Public Transfers.”
In Dominque van de Walle and Kimberley Nead, eds. Public Spending and the Poor:
Theory and Evidence. Baltimore, MD: The Johns Hopkins University Press.
Arguello, Ricardo. 2001. “An Exploration of the Relationship Between Inequality and
Economic Growth. Mimeo, Department of Applied Economics and Management, Cornell
University.
Atkinson, Anthony and Andrea Brandolini. 2001. “Promises and Pitfalls in the Use of
Secondary Data Sets: A Case Study of OECD Income Inequality, Journal of Economic
Literature.
Banerjee, Abhijit and Esther Duflo. 2000. "Inequality and Growth: What can the Data
Say?” Mimeo, MIT.
Barro, Robert, J. and Jong-Wha Lee. 1994. “Sources of Economic Growth.” CarnegieRochester Conference on Public Policy, 40:1-46.
Basu, K. 2001. “Gender and Say: An Intrahousehold Model of Household Behavior.”
CAE Working Paper, Cornell University.
Benhabib, Jess and Aldo Rustichini. 1996. “Social Conflict, Growth and Income
Distribution.” Journal of Economic Growth, 1(1):125-142.
13
Benhabib, Jess and Mark Spiegel. 1994. “Role of Human Capital in Human
Development: Evidence from Aggregate Cross-Country Data.” Journal of Monetary
Economics, 34(9):143-173 October.
Browning, M. and P.A. Chiappori. 1998. “Efficient Intra-Household Allocation: A
General Characterization and Empirical Tests.” Econometrica, 66(6):1241-78.
Buvinic, Mayra and Geeta Rao Gupta. 1997. “Female Headed Households and female
Maintained families: Are they Worth Targeting to Reduce Poverty in Developing
Countries?” Economic Development and Cultural Change, 45(2):259-280.
Case, Anne. 2001a. “The Primacy of Education.” Working Paper, Research Program in
Development Studies, Princeton University.
Case, Anne. 2001b. “Does Money Protect Health Status? Evidence from South African
Pensions.” Working Paper. Research Program in Development Studies, Princeton
University.
Chattopadhyay, Raghabendra and Esther Duflo. 2001. “Women as Policy makers:
Evidence from an India-Wide Randomized Policy Experiment.” Working Paper.
Massachusetts Institute of Technology.
Cohen, Daniel and Marcel Soto. 2001. “Growth and Human Capital: Good Data, Good
Results.” Technical Paper no. 179. OECD Development Centre, Paris.
Dollar, David and Roberta Gatti. 1999. “Gender Inequality, Income and Growth: Are
Good Times Good for Women?” World Bank, Washington D.C.
www.worldbank.org/gender/prr/dg.pdf
Duflo, Esther. 2000. “Grandmothers and Granddaughters” Old Age Pension and Intrahousehold Allocation in South Africa.” Working Paper. Massachusetts Institute of
Technology.
Easterly, William. 2001. The Elusive Quest for Growth, Cambridge, MA: MIT Press.
Easterly, William. 2001a. “Inequality Does Cause Underdevelopment: New Evidence
From Commodity Endowments, Middle Class Share, and Other Determinants of Per
Capita Income.” Mimeo, Center for Global Development—Institute for International
Economics.
Filmer, Deon. 1999. “The Structure of Social Disparities in Education: Gender and
Wealth.” World Bank, Washington, D.C. www.worldbank.org/gender/prr/filmer.pdf
Forbes, Kristin. 2000. “A Reassessment of the Relationship Between Inequality and
Growth.” American Economic Review, 90(4):869-887.
14
Foster, James, Joel Greer and Erik Thorbecke. 1984. “A Class of Decomposable Poverty
Measures.” Econometrica, 52:761-766.
Ghosh, Suman and Ravi Kanbur. 2002. “Male Wages and Female Welfare: Private
Markets, Public Goods and Intrahouseold Inequality.” Cornell University working paper,
http://www.arts.cornell.edu/poverty/kanbur/Ghosh-Kanbur%20.pdf
Haddad, Lawrence and John Hoddinott. 1994. “Women’s Income and Boy-Girl
Anthropometric Status in Cote D’Ivoire.” World Development, 22(4):543-553.
Haddad, L. and R. Kanbur. 1990. “How Serious is the Neglect of Intrahousehold
Inequality?” Economic Journal, 100(402):866-881.
Haddad, L. and R. Kanbur. 1993. “The Value of Intrahousehold Survey Data for Agebased Nutritional Targeting.” Annales d’Economie et de Statistique, 29:65-81.
Hoff, Karla and Andrew Lyon. 1995. "Non-leaky Buckets: Optimal Redistributive
Taxation and Agency Costs." Journal of Public Economics, 58:365-390.
Ilahi, Nadeem. 1999. “Gender and Allocation of Adult Time: What have we Learnt from
the Empirical Literature?” World Bank, Washington D.C.
www.worldbank.org/gender/prr/wp13.pdf
Jejeebhoy, Shireen, J. 1995. Womens’ Education, Autonomy, and Reproductive
Behavior: Experience from Developing Countries. New York: Oxford University Press.
Judson, Ruth. 1996. “Do Low Human Capital Coefficients Make Sense? A Puzzle and
Some Answers.” Board of Governors of the federal Reserve System. Finance and
Economics Discussion Series 96-13, March.
Kanbur, Ravi. 1987. “Measurement and Alleviation of Poverty: With an Application to
the Effetcs of Macroeconomic Adjustment.” I.M.F. Staff Papers 36:60-85.
Kanbur, Ravi. 2000. “Income Distribution and Development,” In Anthony Atkinson
and Francois Bourguignon, eds. Handbook of Income Distribution: Volume I,
Amsterdam: Elsevier.
Kanbur, Ravi (Editor). 2001. “Qualitative and Quantitative Poverty Appraisal:
Complementrities, Tensions and the Way Forward.” Cornell University.
www.people.cornell.edu/pages/sk145/papers/QQZ.pdf
Kanbur, Ravi. 2002. “Economics, Social Science and Development.” World
Development, forthcoming.
Kanbur, Ravi and Lawrence Haddad. 1994. “Are Better off Households More Unequal
or Less Unequal?” Oxford Economic Papers, 46:445-458.
15
Kanbur, Ravi and Nora Lustig. 2000. “Why is Inequality Back on the Agenda?
Proceedigns of the Annual World Bank Conference on Development Economics 1999.
World Bank, Washington, D.C.
Klasen, Stephan. 1999. Does Gender Inequality Reduce Growth and Development?
Evidence from Cross-Country Regressions.” World Bank, Washington, D.C.
www.worldbank.org/gender/prr/klasen.pdf
Klenow, Peter and Andres Rodriguez-Clare. 1997. “The Neoclassical Revival in Growth
Economics: Has it Gone Too Far?” In Ben Bernanke and Julio Rotemberg, eds., NBER
Macroeconomics Annual 1997. Cambridge, MA: MIT Press.
Knowles, Stephen, Paula K. Lorgelly and P. Dorian. 2002. “Are Educational Gender
Gaps a Brake on Economic Development? Some Cross-Country Empirical Evidence.”
Oxford Economic Papers.
Lampieti, Julian A. and Linda Stalker. 2000. “Consumption Expenditure and Female
Poverty: A Review of the Evidence.” World Bank, Washington D.C.
www.worldbank.org/gender/prr/wp11.pdf.
Li, Hongyi, Lyn Squire and Heng-fu Zou. 1998. “Explaining International and
Intertemporal Variations in Income Inequality.” Economic Journal, 108:26-43.
Li, Hongyi, and Heng-fu Zou. 1998. “Income Inequality is not Harmful to Growth:
Theory and Evidence.” Review of Development Economics, 2:318-334.
Lundberg, Shelley, Robert A. Pollak, and Terence J. Wales. 1997. “Do Husbands and
Wives Pool Their Resources? Evidence from the United Kingdom Child Benefit.”
Journal of Human Resources, 32(3):463-480.
Mammen, Kristin and Christina Paxson. 2000. “Women’s Work and Economic
Development.” Journal of Economic Perspectives, 14(4):141-164.
Manser, M and M. Brown. 1980. “Marriage and household decisionmaking: A bargaining
analysis.” International Economic Review, 21(1):31-44.
Morrissey, Oliver, Jennifer Mbabazi and Chris Milner. 2002. “The Fragility of
Empirical Links Between Inequality, Trade Liberalization and Growth.” Working Paper,
University of Nottingham.
Persson, Torsten and Guido Tabellini. 1994. Is Inequality Harmful for Growth?”
American Economic Review, 84(3):600-621.
16
Quisumbing, Agnes R., Lawrence Haddad and Christine L. Pena. 2000. “Are Women
Overrepresented Among the Poor? Poverty Meausres and Dominance Anaalysis for Ten
Developing Countries.” Mimeo, International Food Policy Research Institute,
Washington, D.C.
Quisumbing, Agnes R. and John Maluccio. 1999. “Intrahousehold Allocation and
Gender Relations: New Empirical Evidence.” World Bank, Washington, D.C.
www.worldbank.org/gender/prr/qm.pdf
Pritchett, L. 1997. “Where has all the Education Gone?” World Bank Policy Research
Working Paper, no. 1581.
Ravallion, Martin and K. Chao. 1989. “Targeted Policies for Poverty Alleviation Under
Imperfect Information.” Journal of Policy Modeling, 11:213-224.
Ravallion, Martin and Binayak Sen. 1994. “Impacts on Rural Poverty of Land-Based
Targeting.” World Development, 22:823-838.
Schultz, T. Paul. 2001. “Why Governments Should Make Investments to Educate
Girls.” Working Paper, Yale University.
Shorrocks, Anthony, F. 1984. “Inequality Decomposition by Population Subgroups.”
Econometrica, 52:1369-1385.
Singh, I.J, L. Squire and J. Strauss. 1986. Agriculture Household Models: Extensions,
Applications and Policy. Baltimore: Johns Hopkins University Press.
Thomas, Duncan. 1997. “Incomes, Expenditures and Health Outcomes: Evidence on
Intrahousehold Resource Allocation.” In Lawrence Haddad, John Hoddinott and Harold
Alderman, eds., Intrahousehold Resource Allocation in Developing countries: Models,
Methods and Policy. Baltimore, MD: The Johns Hopkins Press.
Tzannatos, Zafiris. 1999. “Women and Labor Market Changes: Growth Helps,
Inequalities Hurt and Public Policy Matters.” World Development, 27(3):551-69.
World Bank. 2001. Engendering Development Through Gender Equality in Rights,
Resources, and Voice. New York: Oxford University Press.
17